Skip to contentSkip to conversation

lib: stackdepot: Add support to configure STACK_HASH_SIZE

PATCH Superseded [email protected] 4 participants 9 messages Oct 22, 2020 – Nov 20, 2020

Subsystems (2)

v1 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 lib: stackdepot: Add support to configure STACK_HASH_SIZE

From: Yogesh Lal <[email protected]>

Use STACK_HASH_ORDER_SHIFT to configure STACK_HASH_SIZE.

Aim is to have configurable value for STACK_HASH_SIZE, so depend on use case one can configure it.

One example is of Page Owner, default value of STACK_HASH_SIZE lead stack depot to consume 8MB of static memory. Making it configurable and use lower value helps to enable features like CONFIG_PAGE_OWNER without any significant overhead.

---
 lib/Kconfig      | 9 +++++++++
 lib/stackdepot.c | 3 +--
 2 files changed, 10 insertions(+), 2 deletions(-)
lib/Kconfig +9 -0
index 18d76b6..b3f8259 100644
@@ -651,6 +651,15 @@ config STACKDEPOT
 	bool
 	select STACKTRACE
 
+config STACK_HASH_ORDER_SHIFT
+	int "stack depot hash size (12 => 4KB, 20 => 1024KB)"
+	range 12 20
+	default 20
+	depends on STACKDEPOT
+	help
+	 Select the hash size as a power of 2 for the stackdepot hash table.
+	 Choose a lower value to reduce the memory impact.
+
 config SBITMAP
 	bool
 
lib/stackdepot.c +1 -2
index 2caffc6..413c20b 100644
@@ -142,8 +142,7 @@ static struct stack_record *depot_alloc_stack(unsigned long *entries, int size,
 	return stack;
 }
 
-#define STACK_HASH_ORDER 20
-#define STACK_HASH_SIZE (1L << STACK_HASH_ORDER)
+#define STACK_HASH_SIZE (1L << CONFIG_STACK_HASH_ORDER_SHIFT)
 #define STACK_HASH_MASK (STACK_HASH_SIZE - 1)
 #define STACK_HASH_SEED 0x9747b28c
 
Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation
2.7.4
#1 lore raw

Sorry if this mail corrupts the mail thread or had heavy mangling since I lost this mail from my mailbox so I am sending this mail by web gmail.

Quoted 56 lines · On Thu, Oct 22, 2020 at 10:18 AM <[email protected]> wrote:

From: Yogesh Lal <[email protected]>

Use STACK_HASH_ORDER_SHIFT to configure STACK_HASH_SIZE.

Aim is to have configurable value for STACK_HASH_SIZE, so depend on use case one can configure it.

One example is of Page Owner, default value of STACK_HASH_SIZE lead stack depot to consume 8MB of static memory. Making it configurable and use lower value helps to enable features like CONFIG_PAGE_OWNER without any significant overhead.

---
 lib/Kconfig      | 9 +++++++++
 lib/stackdepot.c | 3 +--
 2 files changed, 10 insertions(+), 2 deletions(-)
lib/Kconfig +9 -0
index 18d76b6..b3f8259 100644
@@ -651,6 +651,15 @@ config STACKDEPOT
        bool
        select STACKTRACE
 
+config STACK_HASH_ORDER_SHIFT
+       int "stack depot hash size (12 => 4KB, 20 => 1024KB)"
+       range 12 20
+       default 20
+       depends on STACKDEPOT
+       help
+        Select the hash size as a power of 2 for the stackdepot hash table.
+        Choose a lower value to reduce the memory impact.
+
 config SBITMAP
        bool
 
lib/stackdepot.c +1 -3
index 2caffc6..413c20b 100644
@@ -142,8 +142,7 @@ static struct stack_record *depot_alloc_stack(unsigned long *entries, int size,
        return stack;
 }
 
-#define STACK_HASH_ORDER 20
-#define STACK_HASH_SIZE (1L << STACK_HASH_ORDER)
+#define STACK_HASH_SIZE (1L << CONFIG_STACK_HASH_ORDER_SHIFT)
 #define STACK_HASH_MASK (STACK_HASH_SIZE - 1)
 #define STACK_HASH_SEED 0x9747b28c
 
--
Quoted 3 lines

QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation 2.7.4

1. When we don't use page_owner, we don't want to waste any memory for stackdepot hash array.
2. When we use page_owner, we want to have reasonable stackdeport hash array

With this configuration, it couldn't meet since we always need to reserve a reasonable size for the array.
Can't we make the hash size as a kernel parameter?
With it, we could use it like this.

1. page_owner=off, stackdepot_stack_hash=0 -> no more wasted memory when we don't use page_owner
2. page_owner=on, stackdepot_stack_hash=8M -> reasonable hash size when we use page_owner.

Signature · 2 lines
Kind regards,
Minchan Kim
#2 lore raw
Quoted 74 lines · Minchan Kim

Sorry if this mail corrupts the mail thread or had heavy mangling since I lost this mail from my mailbox so I am sending this mail by web gmail.

Quoted 56 lines · On Thu, Oct 22, 2020 at 10:18 AM <[email protected]> wrote:

From: Yogesh Lal <[email protected]>

Use STACK_HASH_ORDER_SHIFT to configure STACK_HASH_SIZE.

Aim is to have configurable value for STACK_HASH_SIZE, so depend on use case one can configure it.

One example is of Page Owner, default value of STACK_HASH_SIZE lead stack depot to consume 8MB of static memory. Making it configurable and use lower value helps to enable features like CONFIG_PAGE_OWNER without any significant overhead.

---
 lib/Kconfig      | 9 +++++++++
 lib/stackdepot.c | 3 +--
 2 files changed, 10 insertions(+), 2 deletions(-)
lib/Kconfig +9 -0
index 18d76b6..b3f8259 100644
@@ -651,6 +651,15 @@ config STACKDEPOT
        bool
        select STACKTRACE
 
+config STACK_HASH_ORDER_SHIFT
+       int "stack depot hash size (12 => 4KB, 20 => 1024KB)"
+       range 12 20
+       default 20
+       depends on STACKDEPOT
+       help
+        Select the hash size as a power of 2 for the stackdepot hash table.
+        Choose a lower value to reduce the memory impact.
+
 config SBITMAP
        bool
 
lib/stackdepot.c +1 -3
index 2caffc6..413c20b 100644
@@ -142,8 +142,7 @@ static struct stack_record *depot_alloc_stack(unsigned long *entries, int size,
        return stack;
 }
 
-#define STACK_HASH_ORDER 20
-#define STACK_HASH_SIZE (1L << STACK_HASH_ORDER)
+#define STACK_HASH_SIZE (1L << CONFIG_STACK_HASH_ORDER_SHIFT)
 #define STACK_HASH_MASK (STACK_HASH_SIZE - 1)
 #define STACK_HASH_SEED 0x9747b28c
 
--
Quoted 3 lines

QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation 2.7.4

1. When we don't use page_owner, we don't want to waste any memory for stackdepot hash array.
2. When we use page_owner, we want to have reasonable stackdeport hash array

With this configuration, it couldn't meet since we always need to reserve a reasonable size for the array.
Can't we make the hash size as a kernel parameter?
With it, we could use it like this.

Quoted 6 lines

1. page_owner=off, stackdepot_stack_hash=0 -> no more wasted memory when we don't use page_owner
2. page_owner=on, stackdepot_stack_hash=8M -> reasonable hash size when we use page_owner.

This idea looks fine to me. Andrew and others would like to hear your comments as well on this before implementing.

Thanks,
Vijay

Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a
member of Code Aurora Forum, hosted by The Linux Foundation
#3 lore raw
Quoted 84 lines · Vijayanand Jitta
Quoted 74 lines · Minchan Kim

Sorry if this mail corrupts the mail thread or had heavy mangling since I lost this mail from my mailbox so I am sending this mail by web gmail.

Quoted 56 lines · On Thu, Oct 22, 2020 at 10:18 AM <[email protected]> wrote:

From: Yogesh Lal <[email protected]>

Use STACK_HASH_ORDER_SHIFT to configure STACK_HASH_SIZE.

Aim is to have configurable value for STACK_HASH_SIZE, so depend on use case one can configure it.

One example is of Page Owner, default value of STACK_HASH_SIZE lead stack depot to consume 8MB of static memory. Making it configurable and use lower value helps to enable features like CONFIG_PAGE_OWNER without any significant overhead.

---
 lib/Kconfig      | 9 +++++++++
 lib/stackdepot.c | 3 +--
 2 files changed, 10 insertions(+), 2 deletions(-)
lib/Kconfig +9 -0
index 18d76b6..b3f8259 100644
@@ -651,6 +651,15 @@ config STACKDEPOT
        bool
        select STACKTRACE
 
+config STACK_HASH_ORDER_SHIFT
+       int "stack depot hash size (12 => 4KB, 20 => 1024KB)"
+       range 12 20
+       default 20
+       depends on STACKDEPOT
+       help
+        Select the hash size as a power of 2 for the stackdepot hash table.
+        Choose a lower value to reduce the memory impact.
+
 config SBITMAP
        bool
 
lib/stackdepot.c +1 -3
index 2caffc6..413c20b 100644
@@ -142,8 +142,7 @@ static struct stack_record *depot_alloc_stack(unsigned long *entries, int size,
        return stack;
 }
 
-#define STACK_HASH_ORDER 20
-#define STACK_HASH_SIZE (1L << STACK_HASH_ORDER)
+#define STACK_HASH_SIZE (1L << CONFIG_STACK_HASH_ORDER_SHIFT)
 #define STACK_HASH_MASK (STACK_HASH_SIZE - 1)
 #define STACK_HASH_SEED 0x9747b28c
 
--
Quoted 3 lines

QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation 2.7.4

1. When we don't use page_owner, we don't want to waste any memory for stackdepot hash array.
2. When we use page_owner, we want to have reasonable stackdeport hash array

With this configuration, it couldn't meet since we always need to reserve a reasonable size for the array.
Can't we make the hash size as a kernel parameter?
With it, we could use it like this.

Quoted 6 lines

1. page_owner=off, stackdepot_stack_hash=0 -> no more wasted memory when we don't use page_owner
2. page_owner=on, stackdepot_stack_hash=8M -> reasonable hash size when we use page_owner.

Quoted 6 lines

This idea looks fine to me. Andrew and others would like to hear your comments as well on this before implementing.

Thanks,
Vijay

Awaiting for comments from Andrew and others.

Thanks,
Vijay

Signature · 2 lines
QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a
member of Code Aurora Forum, hosted by The Linux Foundation
maintainer
#4 lore raw
Quoted 7 lines · On Thu, 12 Nov 2020 18:26:24 +0530 Vijayanand Jitta <[email protected]>…
Quoted 7 lines
Quoted 6 lines

1. page_owner=off, stackdepot_stack_hash=0 -> no more wasted memory when we don't use page_owner
2. page_owner=on, stackdepot_stack_hash=8M -> reasonable hash size when we use page_owner.

Quoted 8 lines
Quoted 6 lines

This idea looks fine to me. Andrew and others would like to hear your comments as well on this before implementing.

Thanks,
Vijay

Awaiting for comments from Andrew and others.

I don't actually understand the problem.

What is it about page-owner that causes stackdepot to consume additional memory? As far as I can tell, sizeof(struct stack_record) isn't affected by page-owner?

#5 lore raw
Quoted 16 lines · Andrew Morton
Quoted 14 lines · On Thu, 12 Nov 2020 18:26:24 +0530 Vijayanand Jitta <[email protected]>…
Quoted 13 lines
Quoted 6 lines

1. page_owner=off, stackdepot_stack_hash=0 -> no more wasted memory when we don't use page_owner
2. page_owner=on, stackdepot_stack_hash=8M -> reasonable hash size when we use page_owner.

This idea looks fine to me. Andrew and others would like to hear your comments as well on this before implementing.

Thanks,
Vijay

Quoted 8 lines
Quoted 1 line

Awaiting for comments from Andrew and others.

I don't actually understand the problem.

What is it about page-owner that causes stackdepot to consume additional memory? As far as I can tell, sizeof(struct stack_record) isn't affected by page-owner?

Thing is once we build stackdepot due to the dependency from page_owner, it will consume 8M regardless of using page_owner.

#define STACK_HASH_SIZE (1L << CONFIG_STACK_HASH_ORDER_SHIFT)

static struct stack_record *stack_table[STACK_HASH_SIZE] = {
	[0 ...  STACK_HASH_SIZE - 1] = NULL
};

So if we decide the size option at build time, we should consume the memory anyway regardless of page_owner enabling in runtime.

#6 lore raw
In reply to Minchan Kim
Quoted 77 lines · Minchan Kim

Sorry if this mail corrupts the mail thread or had heavy mangling since I lost this mail from my mailbox so I am sending this mail by web gmail.

Quoted 58 lines · On Thu, Oct 22, 2020 at 10:18 AM <[email protected]> wrote:

From: Yogesh Lal <[email protected]>

Use STACK_HASH_ORDER_SHIFT to configure STACK_HASH_SIZE.

Aim is to have configurable value for STACK_HASH_SIZE, so depend on use case one can configure it.

One example is of Page Owner, default value of STACK_HASH_SIZE lead stack depot to consume 8MB of static memory. Making it configurable and use lower value helps to enable features like CONFIG_PAGE_OWNER without any significant overhead.

---
 lib/Kconfig      | 9 +++++++++
 lib/stackdepot.c | 3 +--
 2 files changed, 10 insertions(+), 2 deletions(-)
lib/Kconfig +7 -0
index 18d76b6..b3f8259 100644
@@ -651,6 +651,15 @@ config STACKDEPOT
        bool
        select STACKTRACE
 
+config STACK_HASH_ORDER_SHIFT
+       int "stack depot hash size (12 => 4KB, 20 => 1024KB)"
+       range 12 20
+       default 20
+       depends on STACKDEPOT
+       help
+        Select the hash size as a power of 2 for the stackdepot hash 

table.

lib/Kconfig +2 -0
+        Choose a lower value to reduce the memory impact.
+
 config SBITMAP
        bool
lib/stackdepot.c +1 -3
index 2caffc6..413c20b 100644
@@ -142,8 +142,7 @@ static struct stack_record *depot_alloc_stack(unsigned 
 long *entries, int size,
        return stack;
 }
 
-#define STACK_HASH_ORDER 20
-#define STACK_HASH_SIZE (1L << STACK_HASH_ORDER)
+#define STACK_HASH_SIZE (1L << CONFIG_STACK_HASH_ORDER_SHIFT)
 #define STACK_HASH_MASK (STACK_HASH_SIZE - 1)
 #define STACK_HASH_SEED 0x9747b28c
 
--
Quoted 4 lines

QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation
2.7.4

1. When we don't use page_owner, we don't want to waste any memory for stackdepot hash array.
2. When we use page_owner, we want to have reasonable stackdeport hash array

With this configuration, it couldn't meet since we always need to reserve a reasonable size for the array.
Can't we make the hash size as a kernel parameter?
With it, we could use it like this.

Quoted 4 lines

1. page_owner=off, stackdepot_stack_hash=0 -> no more wasted memory when we don't use page_owner
2. page_owner=on, stackdepot_stack_hash=8M -> reasonable hash size when we use page_owner.

Seems we have other users like kasan, and dma_buf_ref which we introduced. Also we can't guarantee there will not be any other users for stackdepot, so it's better we not depend on only page owner?

Quoted 6 lines
Signature · 3 lines
-- 
Kind regards,
Minchan Kim
#7 lore raw
Quoted 78 lines · Zhenhua Huang
Quoted 77 lines · Minchan Kim

Sorry if this mail corrupts the mail thread or had heavy mangling since I lost this mail from my mailbox so I am sending this mail by web gmail.

Quoted 58 lines · On Thu, Oct 22, 2020 at 10:18 AM <[email protected]> wrote:

From: Yogesh Lal <[email protected]>

Use STACK_HASH_ORDER_SHIFT to configure STACK_HASH_SIZE.

Aim is to have configurable value for STACK_HASH_SIZE, so depend on use case one can configure it.

One example is of Page Owner, default value of STACK_HASH_SIZE lead stack depot to consume 8MB of static memory. Making it configurable and use lower value helps to enable features like CONFIG_PAGE_OWNER without any significant overhead.

---
 lib/Kconfig      | 9 +++++++++
 lib/stackdepot.c | 3 +--
 2 files changed, 10 insertions(+), 2 deletions(-)
lib/Kconfig +7 -0
index 18d76b6..b3f8259 100644
@@ -651,6 +651,15 @@ config STACKDEPOT
        bool
        select STACKTRACE
 
+config STACK_HASH_ORDER_SHIFT
+       int "stack depot hash size (12 => 4KB, 20 => 1024KB)"
+       range 12 20
+       default 20
+       depends on STACKDEPOT
+       help
+        Select the hash size as a power of 2 for the stackdepot hash 

table.

lib/Kconfig +2 -0
+        Choose a lower value to reduce the memory impact.
+
 config SBITMAP
        bool
lib/stackdepot.c +1 -3
index 2caffc6..413c20b 100644
@@ -142,8 +142,7 @@ static struct stack_record *depot_alloc_stack(unsigned 
 long *entries, int size,
        return stack;
 }
 
-#define STACK_HASH_ORDER 20
-#define STACK_HASH_SIZE (1L << STACK_HASH_ORDER)
+#define STACK_HASH_SIZE (1L << CONFIG_STACK_HASH_ORDER_SHIFT)
 #define STACK_HASH_MASK (STACK_HASH_SIZE - 1)
 #define STACK_HASH_SEED 0x9747b28c
 
--
Quoted 4 lines

QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation
2.7.4

1. When we don't use page_owner, we don't want to waste any memory for stackdepot hash array.
2. When we use page_owner, we want to have reasonable stackdeport hash array

With this configuration, it couldn't meet since we always need to reserve a reasonable size for the array.
Can't we make the hash size as a kernel parameter?
With it, we could use it like this.

Quoted 7 lines
Quoted 4 lines

1. page_owner=off, stackdepot_stack_hash=0 -> no more wasted memory when we don't use page_owner
2. page_owner=on, stackdepot_stack_hash=8M -> reasonable hash size when we use page_owner.

Seems we have other users like kasan, and dma_buf_ref which we introduced. Also we can't guarantee there will not be any other users for stackdepot, so it's better we not depend on only page owner?

I didn't mean to make it page_owner dependent. What I suggested is just to define kernel parameter for stackdeport hash size so admin could override it at right size when we really need it.

#8 lore raw
Quoted 89 lines · Minchan Kim
Quoted 9 lines · Zhenhua Huang
Quoted 8 lines · Minchan Kim

Sorry if this mail corrupts the mail thread or had heavy mangling since I lost this mail from my mailbox so I am sending this mail by web gmail.

Quoted 3 lines · On Thu, Oct 22, 2020 at 10:18 AM <[email protected]> wrote:

From: Yogesh Lal <[email protected]>

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

Use STACK_HASH_ORDER_SHIFT to configure STACK_HASH_SIZE.

Aim is to have configurable value for STACK_HASH_SIZE, so depend on use case one can configure it.

One example is of Page Owner, default value of STACK_HASH_SIZE lead stack depot to consume 8MB of static memory. Making it configurable and use lower value helps to enable features

like

Quoted 18 lines
Quoted 18 lines
Quoted 18 lines

CONFIG_PAGE_OWNER without any significant overhead.

---
 lib/Kconfig      | 9 +++++++++
 lib/stackdepot.c | 3 +--
 2 files changed, 10 insertions(+), 2 deletions(-)
lib/Kconfig +0 -0
index 18d76b6..b3f8259 100644
@@ -651,6 +651,15 @@ config STACKDEPOT
        bool
        select STACKTRACE
 
Quoted 7 lines
Quoted 7 lines

lib/Kconfig config STACKDEPOT

Quoted 7 lines
lib/Kconfig +7 -0
+config STACK_HASH_ORDER_SHIFT
+       int "stack depot hash size (12 => 4KB, 20 => 1024KB)"
+       range 12 20
+       default 20
+       depends on STACKDEPOT
+       help
+        Select the hash size as a power of 2 for the stackdepot

hash

Quoted 11 lines
Quoted 11 lines

lib/stackdepot.c static struct stack_record

Quoted 11 lines

table.

lib/Kconfig +2 -0
+        Choose a lower value to reduce the memory impact.
+
 config SBITMAP
        bool
lib/stackdepot.c +0 -0
index 2caffc6..413c20b 100644
@@ -142,8 +142,7 @@ static struct stack_record

*depot_alloc_stack(unsigned

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines
lib/stackdepot.c +0 -0
 long *entries, int size,
        return stack;
 }
 
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
lib/stackdepot.c +1 -3
-#define STACK_HASH_ORDER 20
-#define STACK_HASH_SIZE (1L << STACK_HASH_ORDER)
+#define STACK_HASH_SIZE (1L << CONFIG_STACK_HASH_ORDER_SHIFT)
 #define STACK_HASH_MASK (STACK_HASH_SIZE - 1)
 #define STACK_HASH_SEED 0x9747b28c
 
--

QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a

member

Quoted 7 lines
Quoted 7 lines
Quoted 3 lines

of Code Aurora Forum, hosted by The Linux Foundation 2.7.4

1. When we don't use page_owner, we don't want to waste any memory for stackdepot hash array.
2. When we use page_owner, we want to have reasonable stackdeport hash

array

Quoted 11 lines
Quoted 10 lines

With this configuration, it couldn't meet since we always need to reserve a reasonable size for the array.
Can't we make the hash size as a kernel parameter?
With it, we could use it like this.

1. page_owner=off, stackdepot_stack_hash=0 -> no more wasted memory when we don't use page_owner
2. page_owner=on, stackdepot_stack_hash=8M -> reasonable hash size when we use page_owner.

Seems we have other users like kasan, and dma_buf_ref which we

Quoted 8 lines

introduced.

Quoted 1 line

Also we can't guarantee there will not be any other users for

stackdepot, so

Quoted 1 line

it's better we not depend on only page owner?

I didn't mean to make it page_owner dependent. What I suggested is just to define kernel parameter for stackdeport hash size so admin could override it at right size when we really need it.

OK, Thanks Minchan for explanation. It's a good idea then, admin needs to consider all users but, especailly when setting it to 0...

Thanks,
Zhenhua